The day after anthropologist Jacob Coxon quit his job over concerns that AI could wipe out humanity by the end of the decade, I met founders and brothers-in-law Rune Qvist and Rajiv Dattani. They think they have a solution that can save us all, or at least help prevent AI agents from running amok within companies.
“AI is getting smarter and smarter and faster. The amazing thing about AI is that as it gets smarter, it doesn’t get easier, it gets harder to deploy and harder to control,” said Kvist, an early Anthropic employee who is married to Dattani’s sister. Mr. Dattani is the former COO of AI safety research organization METR.
The two have launched a startup called Artificial Intelligence Underwriting Company (AIUC), which hopes to bring AI safety to businesses and companies that build AI models and agents. The startup counts Cursor, Lovable, Harvey, and Celebrities as customers.
On Tuesday, AIUC announced a $40 million Series A led by Ribbit Capital with participation from First Harmonic. The company previously closed a $15 million seed round through Nat Friedman’s NFDG fund, along with Ben Mann, co-founder of Emergence, Terrain, and Anthropic, bringing total funding to $55 million.
What caught the attention of this group of A-list investors was AIUC’s attempt to apply familiar cybersecurity models to new AI risks. The company has built a third-party audit and authentication layer for its AI agents.
“Banks, hospitals, governments, and militaries will no longer refuse to deploy AI because the models aren’t smart enough,” Kvist says. “They say no because they’ve made promises to their customers about what the system will or won’t do, and no one can guarantee that at this point.”
Using the widely adopted cybersecurity standard SOC 2 as its muse, AIUC developed a standard called AIUC-1 and a testing service to validate agents against that standard.
To build the standard, AIUC assembled a consortium of approximately 250 security and risk leaders who are agent buyers. “These are the people we meet with every month, and the questions we ask them are: When you buy an agent from someone, what do you look for? What questions do you want to ask? And what do you want them to address?” Dattani told TechCrunch.
That feedback shapes the test. The startup then runs the agent through a battery of about 5,000 tests to see how it performs in scenarios involving jailbreaks, hallucinations, and data leaks. The result is a nearly 100-page report detailing where agents are working safely and reliably and where they are not. Interestingly, AIUC uses AI agents to run tests and uses AI to analyze data. But the final audit is verified by humans, Kvist said.
If this sounds a little familiar, you’re right. Mr. Dattani’s former employer, METR, where he served as chief operating officer from 2024 to 2025 and remains a board member, is conducting similar testing for Frontier Labs, but until recently that work has focused primarily on performance (making sure agents can complete tasks). METR is one of the independent investigative agencies OpenAI used to investigate the Hugface incident.
Anthropic CEO Dario Amodei also recently called on the AI industry to accelerate the pace of frontier development, citing a spike in fraud cases. In his post, Amodei floated the idea of requiring Frontier Laboratories to use embedded third-party evaluators to observe and verify safety, citing METR as one possibility.
While AIUC doesn’t propose integrating itself into customer sites, the overall idea is similar: to give companies the ability to independently assess the safety of their AI agents. “This is a passing grade, this is where you can trust. And this is where there are concerns. You need to be aware of these references before making a purchasing decision,” Dattani said.
If you buy through links in our articles, we may earn a small commission. This does not affect editorial independence.
